Terminal.skills
Use Cases/Produce Automated Marketing Videos at Scale with Remotion

Produce Automated Marketing Videos at Scale with Remotion

Generate hundreds of personalized product videos with dynamic subtitles and thumbnail frames using code-driven video production.

Automation#remotion#video#react#animation#rendering
Works with:claude-codeopenai-codexgemini-clicursor
remotion-video-toolkit
video-subtitles
video-frames
$

The Problem

An e-commerce brand with 350 products needs short promotional videos for each product listing on their website and social media. A freelance video editor charges $75 per video and takes 3 days for a batch of 20. At that rate, covering all 350 products costs $26,250 and takes months. When prices change or new products launch, existing videos become outdated and the entire cycle restarts. The marketing team also wants each video to have burned-in subtitles for accessibility and autoplay-without-sound contexts, plus thumbnail frames for social media previews.

The Solution

Using remotion-video-toolkit to programmatically generate videos from product data and templates, video-subtitles to burn accurate captions into each video, and video-frames to extract optimized thumbnail images, the brand produces all 350 videos in a single overnight batch run.

Step-by-Step Walkthrough

1. Build a parameterized video template in Remotion

Create a React-based video component that accepts product data as props and renders a polished 15-second promotional clip.

Use remotion-video-toolkit to create a 15-second product video template. The template takes props: productName, price, description, imageUrl, and backgroundColor. Animate the product image with a zoom-in entrance at frame 0, display the product name with a typewriter effect at frame 30, show the price with a bounce animation at frame 60, and end with a call-to-action "Shop Now" button at frame 90. Use 1080x1080 for Instagram and 1920x1080 for YouTube.

2. Generate subtitle tracks for each video

Create timed subtitle files from the product descriptions and burn them into the video frames.

Use video-subtitles to generate SRT subtitle files for each product video. The subtitle text comes from the product's short description, timed to match the animation: product name at 0-2s, key feature at 2-5s, price at 5-8s, and call-to-action at 8-10s. Burn the subtitles into the video with a semi-transparent black background, white text, and the brand's font. Process all 350 videos in batch.

3. Extract thumbnail frames at the most visually appealing moment

Pull optimized still frames from each video for use as social media preview images and product listing thumbnails.

Use video-frames to extract thumbnail images from each of the 350 rendered videos. Capture frames at the 3-second mark (when the product image is fully zoomed in and the name is displayed). Export as both 1080x1080 PNG for Instagram and 1280x720 JPEG for YouTube thumbnails. Apply a slight sharpening filter to compensate for video compression artifacts in the still frame.

4. Batch render all 350 product videos

Feed the product catalog CSV into Remotion's batch renderer and produce all videos overnight.

Use remotion-video-toolkit to batch render all 350 product videos. Read product data from /data/products.csv. For each product, render both Instagram (1080x1080) and YouTube (1920x1080) versions. Use 4 concurrent Remotion render processes. Output to /output/videos/{product_slug}/. Estimated render time: 45 seconds per video, total batch: ~4.4 hours. Log progress and any failures to /output/render_log.csv.

The batch renderer streams progress to the console and writes a final summary:

remotion render --batch /data/products.csv --concurrency 4

[worker-1] Rendering organic-cotton-tee (1/350)... 1080x1080 done (38s), 1920x1080 done (42s)
[worker-2] Rendering ceramic-pour-over (2/350)...  1080x1080 done (41s), 1920x1080 done (44s)
[worker-3] Rendering bamboo-desk-mat (3/350)...    1080x1080 done (36s), 1920x1080 done (40s)
[worker-4] Rendering walnut-knife-set (4/350)...   1080x1080 done (39s), 1920x1080 done (43s)
...
[worker-2] Rendering linen-napkin-set (350/350)... 1080x1080 done (37s), 1920x1080 done (41s)

Batch complete:
  Products rendered:  350
  Videos produced:    700 (350 Instagram + 350 YouTube)
  Subtitled versions: 700
  Thumbnails:         700
  Total render time:  4h 12m
  Failures:           0
  Output size:        28.4 GB
  Log:                /output/render_log.csv

Real-World Example

The e-commerce brand ran the first batch on a Friday night. By Saturday morning, all 350 products had both Instagram and YouTube video variants, subtitle-burned versions for accessibility, and optimized thumbnail frames. Total render time was 4 hours and 12 minutes on a single 8-core machine. When a seasonal sale required updating prices on 80 products, they edited the CSV, re-ran only the changed rows, and had updated videos in 90 minutes. The project that would have cost $26,250 and taken months with a freelancer was completed for the cost of compute time, and new products get their videos the same day they are added to the catalog.

Tips

  • Keep product images at 2x the video resolution (2160x2160 for 1080x1080 video) to allow Remotion's zoom animation to stay sharp without pixelation.
  • Render at 30fps for social media. Most platforms re-encode uploads anyway, and 60fps doubles render time with no visible benefit on 15-second clips.
  • Use a hash of the product data row as a cache key. When re-running the batch, skip any product whose hash has not changed since the last render to cut re-run times dramatically.
  • Test the video template with 5 products before running the full 350. Edge cases like extra-long product names or missing images are cheaper to catch early.